Causality

Overview

This area tracks structural causal models, interventions, counterfactual reasoning, and how causal ideas support trustworthy AI research.

How can causal abstractions make agent behavior more interpretable and more robust?
Which causal tools matter most for sequential decision-making under intervention?
Where do causal explanations complement or outperform post hoc interpretability?
What assumptions let structural models guide interventions rather than merely fit associations?
How should actual-cause claims be represented when the domain is an explicit action history?
Can temporal structural equation models support computation-like causal reasoning without collapsing into acyclic SEMs?
When does SCM notation support moral-cognition modeling, and when do added psychological operators stop being ordinary causal structure?
When does a policy-value or treatment-effect claim require intervention semantics rather than predictive association?
Which causal assumptions are being spent when offline or adaptive decision data are reused for policy learning?
When can an agent causally alter the feedback process used to define or learn its objective?
When does an intervention in a game keep policies fixed, and when must the model recompute rational strategic responses?
When can counterfactual distributions be simulated from a structural causal model rather than derived analytically?